import warnings
warnings.simplefilter('ignore')
import pandas as pd
import numpy as np
import plotly.graph_objects as go
import plotly.express as px

anime_df=pd.read_csv("anime-dataset-2023.csv")
pd.set_option('display.max_columns', None)

anime_df

	anime_id	Name	English name	Other name	Score	Genres	Synopsis	Type	Episodes	Aired	Premiered	Status	Producers	Licensors	Studios	Source	Duration	Rating	Rank	Popularity	Favorites	Scored By	Members	Image URL
0	1	Cowboy Bebop	Cowboy Bebop	カウボーイビバップ	8.75	Action, Award Winning, Sci-Fi	Crime is timeless. By the year 2071, humanity ...	TV	26.0	Apr 3, 1998 to Apr 24, 1999	spring 1998	Finished Airing	Bandai Visual	Funimation, Bandai Entertainment	Sunrise	Original	24 min per ep	R - 17+ (violence & profanity)	41.0	43	78525	914193.0	1771505	https://cdn.myanimelist.net/images/anime/4/196...
1	5	Cowboy Bebop: Tengoku no Tobira	Cowboy Bebop: The Movie	カウボーイビバップ 天国の扉	8.38	Action, Sci-Fi	Another day, another bounty—such is the life o...	Movie	1.0	Sep 1, 2001	UNKNOWN	Finished Airing	Sunrise, Bandai Visual	Sony Pictures Entertainment	Bones	Original	1 hr 55 min	R - 17+ (violence & profanity)	189.0	602	1448	206248.0	360978	https://cdn.myanimelist.net/images/anime/1439/...
2	6	Trigun	Trigun	トライガン	8.22	Action, Adventure, Sci-Fi	Vash the Stampede is the man with a $$60,000,0...	TV	26.0	Apr 1, 1998 to Sep 30, 1998	spring 1998	Finished Airing	Victor Entertainment	Funimation, Geneon Entertainment USA	Madhouse	Manga	24 min per ep	PG-13 - Teens 13 or older	328.0	246	15035	356739.0	727252	https://cdn.myanimelist.net/images/anime/7/203...
3	7	Witch Hunter Robin	Witch Hunter Robin	Witch Hunter ROBIN (ウイッチハンターロビン)	7.25	Action, Drama, Mystery, Supernatural	Robin Sena is a powerful craft user drafted in...	TV	26.0	Jul 3, 2002 to Dec 25, 2002	summer 2002	Finished Airing	Bandai Visual, Dentsu, Victor Entertainment, T...	Funimation, Bandai Entertainment	Sunrise	Original	25 min per ep	PG-13 - Teens 13 or older	2764.0	1795	613	42829.0	111931	https://cdn.myanimelist.net/images/anime/10/19...
4	8	Bouken Ou Beet	Beet the Vandel Buster	冒険王ビィト	6.94	Adventure, Fantasy, Supernatural	It is the dark century and the people are suff...	TV	52.0	Sep 30, 2004 to Sep 29, 2005	fall 2004	Finished Airing	TV Tokyo, Dentsu	Illumitoon Entertainment	Toei Animation	Manga	23 min per ep	PG - Children	4240.0	5126	14	6413.0	15001	https://cdn.myanimelist.net/images/anime/7/215...
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
24900	55731	Wu Nao Monu	UNKNOWN	无脑魔女	UNKNOWN	Comedy, Fantasy, Slice of Life	No description available for this anime.	ONA	15.0	Jul 4, 2023 to ?	UNKNOWN	Not yet aired	UNKNOWN	UNKNOWN	UNKNOWN	Web manga	Unknown	PG-13 - Teens 13 or older	UNKNOWN	24723	0	UNKNOWN	0	https://cdn.myanimelist.net/images/anime/1386/...
24901	55732	Bu Xing Si: Yuan Qi	Blader Soul	捕星司·源起	UNKNOWN	Action, Adventure, Fantasy	No description available for this anime.	ONA	18.0	Jul 27, 2023 to ?	UNKNOWN	Not yet aired	UNKNOWN	UNKNOWN	UNKNOWN	Web novel	Unknown	PG-13 - Teens 13 or older	0.0	0	0	UNKNOWN	0	https://cdn.myanimelist.net/images/anime/1383/...
24902	55733	Di Yi Xulie	The First Order	第一序列	UNKNOWN	Action, Adventure, Fantasy, Sci-Fi	No description available for this anime.	ONA	16.0	Jul 19, 2023 to ?	UNKNOWN	Finished Airing	UNKNOWN	UNKNOWN	UNKNOWN	Web novel	Unknown	PG-13 - Teens 13 or older	0.0	0	0	UNKNOWN	0	https://cdn.myanimelist.net/images/anime/1130/...
24903	55734	Bokura no Saishuu Sensou	UNKNOWN	僕らの最終戦争	UNKNOWN	UNKNOWN	A music video for the song Bokura no Saishuu S...	Music	1.0	Apr 23, 2022	UNKNOWN	Finished Airing	UNKNOWN	UNKNOWN	UNKNOWN	Original	3 min	PG-13 - Teens 13 or older	0.0	0	0	UNKNOWN	0	https://cdn.myanimelist.net/images/anime/1931/...
24904	55735	Shijuuku Nichi	UNKNOWN	四十九日	UNKNOWN	UNKNOWN	A music video for the song Shijuuku Nichi by S...	Music	1.0	Sep 5, 2022	UNKNOWN	Finished Airing	UNKNOWN	UNKNOWN	UNKNOWN	Original	3 min	PG-13 - Teens 13 or older	0.0	0	0	UNKNOWN	0	https://cdn.myanimelist.net/images/anime/1902/...

24905 rows × 24 columns

Filter dataset

as could be seen Anime dataset has some columns that don't give any useful information for analysys or double information from other columns(English name,other name,Synopsis,Image URL)

anime_df=anime_df.drop(columns=['Image URL'], axis=0)
anime_df=anime_df.drop(columns=['Licensors'], axis=0)
anime_df=anime_df.drop(columns=['Producers'], axis=0)
anime_df=anime_df.drop(columns=['English name'], axis=0)
anime_df=anime_df.drop(columns=['Other name'], axis=0)
anime_df=anime_df.drop(columns=['Synopsis'], axis=0)

print(anime_df['Score'].describe())

count       24905
unique        567
top       UNKNOWN
freq         9213
Name: Score, dtype: object

Let's filter data so we can get more information about Score

anime_df['Score'] = anime_df['Score'].replace('UNKNOWN', np.nan)
anime_df['Score'] = pd.to_numeric(anime_df['Score'], errors='coerce')
anime_df['Score'].isna().sum()

np.int64(9213)

anime_df=anime_df.dropna(subset='Score')

print(anime_df['Score'].describe())

count    15692.000000
mean         6.380890
std          0.928696
min          1.850000
25%          5.730000
50%          6.390000
75%          7.060000
max          9.100000
Name: Score, dtype: float64

We get data description, I had to delete 9k rows, but I want to work with clear data, so such manipulation is appropriate

Let's check rating column and reform data in numeric format.

anime_df['Rating'].unique()

array(['R - 17+ (violence & profanity)', 'PG-13 - Teens 13 or older',
       'PG - Children', 'R+ - Mild Nudity', 'G - All Ages', 'Rx - Hentai',
       'UNKNOWN'], dtype=object)

rating_mapping = {
    'Rx - Hentai': 21,
    'R+ - Mild Nudity': 18,
    'R - 17+ (violence & profanity)': 17,
    'PG-13 - Teens 13 or older': 13,
    'PG - Children': 10,
    'G - All Ages': 0
}

anime_df['Rating']=anime_df['Rating'].replace(rating_mapping)
anime_df['Rating'] = anime_df['Rating'].replace('UNKNOWN', np.nan)
anime_df=anime_df.dropna(subset='Rating')
anime_df['Rating']=anime_df['Rating'].astype(int)

anime_df['Rating'].describe()

count    15598.000000
mean        10.956661
std          6.897628
min          0.000000
25%         10.000000
50%         13.000000
75%         13.000000
max         21.000000
Name: Rating, dtype: float64

Now we can look at Duration column, but we also have to reformat type into numeric.

anime_df['Duration']=anime_df['Duration'].replace('min',' ')

import re
def convert_to_minutes(duration):
    hours = re.search(r'(\d+)\s*hr', duration)
    minutes = re.search(r'(\d+)\s*min', duration) 
    
    total_minutes = 0
    if hours:
        total_minutes += int(hours.group(1)) * 60
    if minutes:
        total_minutes += int(minutes.group(1))
    return total_minutes

    anime_df['Duration'] = anime_df['Duration'].apply(convert_to_minutes)

Let's convert the Status data into numeric.

anime_df['Status'].value_counts()

Status
Finished Airing     15451
Currently Airing      147
Name: count, dtype: int64

anime_df['Status']=anime_df['Status'].replace('Finished Airing',1)
anime_df['Status']=anime_df['Status'].replace('Currently Airing',2)

I want my data about duration to be presented in months, so I write a function to transform data.

from datetime import datetime
def parse_date(date_str):
    for fmt in ("%b %d, %Y", "%b %Y"):
        try:
            return pd.to_datetime(date_str, format=fmt)
        except ValueError:
            continue
    return pd.NaT
def calculate_months(aired_range):
    try:
        start, end = aired_range.split(' to ')
        start_date = parse_date(start)
        end_date = parse_date(end)
        if pd.notna(start_date) and pd.notna(end_date):
            return (end_date.year - start_date.year) * 12 + (end_date.month - start_date.month)
        return None
    except ValueError:
        return None
anime_df['Aired'] = anime_df['Aired'].apply(calculate_months)

As could be seen from the Genres column there are several genres for each row, but this makes data really hard to identify and make graphs, do I will take the first genre from the each row , because first mentions genre is the major for anime.

I also get rid of UNKNOWN values.

anime_df['Genres'] = anime_df['Genres'].apply(lambda x: x.split(',')[0].strip())

anime_df = anime_df[anime_df['Genres'] != 'UNKNOWN']

anime_df['Type'] = anime_df['Type'].replace('UNKNOWN', np.nan)

As there are a lot of genres that have a really small amount of representors, so I put some Genres into 'Other' group.

genre_counts=anime_df['Genres'].value_counts()

common_genres = genre_counts[genre_counts > 200]  

anime_df['Filtered Genres'] = anime_df['Genres'].apply(lambda x: x if x in common_genres.index else 'Other')

filtered_genre_counts = anime_df['Filtered Genres'].value_counts()

othergenres = genre_counts[genre_counts <= 200]

fig = px.pie(
    names=filtered_genre_counts.index,
    values=filtered_genre_counts.values,
    title=' Major genres pie chart',
    width=800,  
    height=800  
)

fig.update_traces(
    textposition='inside',
    textinfo='percent+label' 
)
fig.update_layout(
    title_font_size=24,
    legend_font_size=16  
)
fig.write_image("MJ.png") 
fig.show()

Let's look closer at the 'Other' group chart

fig1=px.pie(names=othergenres.index,
            values=othergenres.values,
            title='Minor genres pie chart',
            width=800,  
            height=800 )
fig1.write_image("minor_genres.png") 
fig1.show()

avg_score=anime_df.groupby('Genres') ['Score'].mean()
df = avg_score.reset_index()
print(df)

           Genres     Score
0          Action  6.679460
1       Adventure  6.569881
2     Avant Garde  5.095424
3   Award Winning  7.153308
4       Boys Love  6.509141
5          Comedy  6.435754
6           Drama  6.739659
7           Ecchi  6.056600
8         Erotica  5.022000
9         Fantasy  6.141020
10     Girls Love  6.282000
11        Gourmet  6.620000
12         Hentai  6.119391
13         Horror  5.769016
14        Mystery  6.728507
15        Romance  6.357478
16         Sci-Fi  6.057866
17  Slice of Life  6.178711
18         Sports  6.927021
19   Supernatural  6.199322
20       Suspense  6.610000

Let's check which genre has the highest score.

fig = px.bar(df, x="Genres", y="Score", title="Genre-Score chart",
            width=800,
            height=800)
fig.update_layout(yaxis=dict(range=[0, 10],
                            tick0=0,
                            dtick=0.7,
                            ))
fig.write_image("genre-score.png") 
fig.show()

anime_df['Premiered'] = anime_df['Premiered'].replace('UNKNOWN', np.nan)
anime_df = anime_df.dropna(subset=['Premiered'])
anime_df['Premiered'] = anime_df['Premiered'].str.extract(r"(\d{4})")

genre_counts = anime_df.groupby(['Premiered', 'Genres']).size().reset_index(name='Count')

total_shows_per_year = genre_counts.groupby('Premiered')['Count'].transform('sum')

genre_counts['Proportion'] = genre_counts['Count'] / total_shows_per_year

heatmap_data = genre_counts.pivot(index='Genres', columns='Premiered', values='Proportion').fillna(0)

fig = px.imshow(
    heatmap_data,
    x=heatmap_data.columns,  
    y=heatmap_data.index,    
    color_continuous_scale="plotly3",
    labels=dict(x="Premiered Year", y="Genres", color="Proportion"),
    title="Proportion of anime genres over time" 
)

fig.update_layout(
    xaxis=dict(tickangle=45),
)

fig.write_image("proportion.png") 
fig.show()

year_counts = anime_df['Premiered'].value_counts()
sorted_years = sorted(year_counts.index)

fig = go.Figure(data=go.Bar(
    x=sorted_years,
    y=year_counts[sorted_years], 
))

fig.update_layout(
    title='Number of Animes Premiered by Year',
    xaxis_title='Year',
    yaxis_title='Number',
    title_font=dict(size=20),
)
fig.write_image("number_animes.png") 
fig.show()

anime_df['Episodes'] = anime_df['Episodes'].replace('UNKNOWN', np.nan)
anime_df = anime_df.dropna(subset=['Episodes'])
anime_df['Episodes'] = anime_df['Episodes'].astype(float)
anime_df['Episodes'] = anime_df['Episodes'].astype(int)
episodes_groups=anime_df['Episodes'].value_counts()
episode_distribution = anime_df['Episodes'].value_counts()
print(episode_distribution)

Episodes
12     1405
13      639
26      402
24      227
25      176
       ... 
128       1
117       1
172       1
500       1
98        1
Name: count, Length: 175, dtype: int64

anime_df['Rank'] = anime_df['Rank'].replace('UNKNOWN', np.nan)
anime_df=anime_df.dropna(subset='Rank')

As the rank system is described in 10000 let's make it in decimal gradation.

anime_df['Rank']=anime_df['Rank'].astype(float)
min_value = anime_df['Rank'].min()
max_value = anime_df['Rank'].max()
anime_df['Rank'] = (anime_df['Rank'] - min_value) / (max_value - min_value) * 10
anime_df['Rank'] = np.ceil(anime_df['Rank']).astype(int)

correlation_matrix = anime_df[['Score', 'Popularity', 'Rank']].corr()

fig = go.Figure(data=go.Heatmap(
    z=correlation_matrix.values,
    x=correlation_matrix.columns,
    y=correlation_matrix.index,
    colorscale='plasma',
    zmin=-1, zmax=1,
    colorbar=dict(title="Correlation")
))

fig.update_layout(
    title="Correlation Matrix for Score, Popularity, and Rank",
    title_font_size=18,
    font=dict(size=12),
    xaxis=dict(tickangle=0)
)
fig.write_image("correlation.png") 
fig.show()

Hypotesis:

Source of anime affects it's score and popularity , so major genres also rely on that sources that bring the highest popularity-score ratio.

Firstly let's look how sources affect the animes in general.

filtered_data = anime_df[anime_df["Source"] != "Unknown"]
grouped_data = filtered_data.groupby("Source").agg(
    avg_score=("Score", "mean"),
    avg_popularity=("Popularity", "mean")
).reset_index()


fig = px.scatter(
    grouped_data,
    x="avg_popularity",
    y="avg_score",
    size="avg_popularity",
    color="Source",
    hover_name="Source",
    labels={"avg_popularity": "Average Popularity", "avg_score": "Average Score"},
    title="Dependency of Score and Popularity on Anime Source"
)

fig.update_layout(
    xaxis_title="Average Popularity",
    yaxis_title="Average Score",
)
fig.write_image("scatter_1.png") 
fig.show()

Now I want to look at each genre to chek if source really affect genres.

statistics = anime_df.groupby(["Source", "Genres"]).agg(
    avg_score=("Score", "mean"),
    avg_popularity=("Popularity", "mean"),
    anime_count=("Score", "count")
).reset_index()

fig = px.scatter(
    statistics,
    x="avg_popularity",
    y="avg_score",
    size="anime_count",
    color="Source",
    facet_col="Genres", 
    facet_col_wrap=4, 
    hover_data=["Source", "avg_popularity", "avg_score", "anime_count"],
    labels={
        "avg_popularity": "Average Popularity",
        "avg_score": "Average Score",
        "anime_count": "Number of Anime",
        "Genre": "Genre" 
    },
    title="Popularity-Score for Genres by Source",
    size_max=40  
)

fig.update_layout(
    width=1200, 
    height=1000,   
    font=dict(size=10),  

)

fig.show()
fig.write_image("scatter_2.png") 

As could be seen from "Major genres pie chart" and "Popularity-Score for Genres by Source" major genres mostly have Manga as most used source and "Dependency of Score and Popularity on Anime Source" Manga leads by average score, so the hypotesis is true.